- Memorization of feature interactions through a wide set of cross-product feature transformations are effective and interpretable, while generalization requires more feature engineering effort.
- Deep neural networks can generalize better to unseen feature combinations through low-dimensional dense embeddings learned for the sparse features.
- In the paper, we present Wide & Deep learning—jointly trained wide linear models and deep neural networks—to combine the benefits of memorization and generalization for recommender systems.
- Deep neural networks with embeddings can over-generalize and recommend less relevant items when the user-item interactions are sparse and high-rank.
- Abstract: Generalized linear models with nonlinear feature transformations are widely used for large-scale regression and classification problems with sparse inputs.
Read the full article, click here.
@graphific: “Combining wide linear features with deep neural nets – Paper: Code:”
One hundred percent of your contribution will fund improvements and new initiatives to benefit arXiv’s global scientific community. Please join the Simons Foundation and our generous member organizations and research labs in supporting arXiv.
[1606.07792] Wide & Deep Learning for Recommender Systems